paper 105 comparison of scale model measurements and 3d

“SMART RIVERS 2015”

Buenos Aires, Argentina, 7-11 September 2015

SMART RIVERS 2015 (www.pianc.org.ar/sr2015) Paper 137 - /8

Paper 105 - Comparison of Scale Model Measurements and 3D CFD Simulations of Loss Coefficients and Flow Patterns for Lock

Levelling Systems

VAN DER VEN, P.P.D., VAN VELZEN, G., O’MAHONEY, T.S.D., DE LOOR, A. Deltares, Hydraulic Engineering, Delft, The Netherlands

Email (1st author): [email protected]

ABSTRACT: The discharge coefficient is a determinative parameter in the prediction of forces on ships during in a lock chamber during levelling. The application of CFD to establish this coefficient a priori has been validated using a scale model of a lock gate with gate openings. Various tests have been executed with accurate results. To allow a further validation of CFD the measurements include the registration of velocities at eighty points downstream of the lock gate. CFD shows good agreement with these measurements, although the dissipation of energy in the flow field downstream of the lock gate is found to be overestimated.

1 INTRODUCTION

1.1 The need to calculate forces during levelling

The purpose of a lock is to allow ship traffic to pass between different bodies of water. The lock’s operation is primarily aimed at minimizing the delay caused by the lock. The levelling process should therefore be completed as quickly as possible, while hawser forces during the levelling must remain within safety limits. To design a lock that levels both quickly and safely is an engineering challenge.

Several contributors to the force acting on a vessel during levelling can be identified.

The flow of water through the levelling system may be concentrated as a filling jet, possibly directed towards the vessel’s hull.

The decrease of discharge along the length of the lock causes a water level slope, resulting in a force on the ship directed towards the active lock head.

The change of discharge in time causes a translatory wave that induces an oscillating force on the ship.

A fourth contribution is found due to the friction of the hull and the flow around the vessel.

Finally, a fifth contribution is found in the case of a density difference between the lock chamber and the approach harbour, causing a density driven flow.

The force contributors described depend, among

others things, on the vessels dimensions, the keel clearance and the distance between the ship and the active lock head. But most importantly, the forces are dependent of the discharge through the levelling system. The discharge, in turn, depends on the cross section and the discharge coefficient of the levelling system: in case of gate openings, the discharge can be described by the following formula.

√

In which Q discharge [m3/s], µ discharge coefficient [-], A the cross-sectional under the door valve, [m2], g gravitational constant [m/s2] and Δh water level difference [m].

1.2 Lockfill

The simulation of the levelling process and the calculation of the longitudinal forces on a vessel have often been done using the calculation program LOCKFILL, developed by Deltares (see De Loor et al. (2013) and the LOCKFILL manual).

This computationally fast one-dimensional model, using cross-sectional averaged quantities, allows various levelling systems through the lock heads. LOCKFILL computes the water level variation and flow in the lock chamber as a function of time, starting from an initial water level difference across

http://www.pianc.org.ar/sr2015




the lock head. An important variable in this computation is the discharge coefficient introduced above.

Subsequently, LOCKFILL computes the five force contributors. The total longitudinal force follows from the summation of these components.

LOCKFILL has been extensively validated with data of existing shipping locks in the Netherlands. It has recently been applied in the evaluation of many existing locks as part of the project RINK (risk assessment of hydraulic structures).

Per July 2015, LOCKFILL has been made freely available via oss.deltares.nl.

1.3 Determining the discharge coefficient of levelling systems

The discharge coefficient or, equivalently, the loss coefficient, follows from the geometry of the levelling system and is not easily determined a priori. Details of the geometry are most important, e.g. widening of the door openings, the breaking beam orientation or the shape of the culverts. Literature is often found insufficient to determine loss coefficients for a specific design and when used, results should be interpreted with caution and with a large error bound.

In the past, the discharge coefficient was determined more accurately by executing scale model measurements. These studies now comprise a small library of coefficients that can be used as reference. This library is, however, very limited and cannot provide a reference for designs that vary greatly from those already tested. Recent technological innovations which are presently being developed, such as composite lock gates, are thus not included. The execution of scale model tests for the determination of discharge coefficients is often quite expensive.

In order to provide a way to determine discharge coefficients, particularly early in the design process, the application of Computaionla Fluid Dynamics (CFD) is very promising. Measurement data that can be used for the validation of CFD in such an application are available but these tests do not provide a detailed flow pattern or velocity measurements which are needed to more fully validate a 3D CFD model. The current study provides this data specifically for the purpose of CFD validation.

1.4 Research question

The research question can thus be described as follows: to allow the validation of CFD in the determination of discharge coefficients of lock gate openings, measurements must be performed that provide sufficiently accurate data.

2 SCALE MODEL EXPERIMENTS

2.1 Test approach

A scale model has been made of a lock gate with openings in the gate. A stationary flow situation has been by using a pump on a constant discharge; the water level difference will set as a result. When the equilibrium has been reached, water levels and discharge are registered, from which the discharge coefficient is calculated.

2.2 Test faclilty

The model is placed in the Easten Scheldt flume at Deltares, having a length of 55 m, width of 1.0 m and a maximum water level of 1.2 m. The flume has glass walls on both sides, which limits the wall frition and allows easy visualization of the flow pattern with ink.

The flume is equipped with a pump on one side. Possible large scale eddies at the inlet are broken up by a resistance screen in order to obtain a uniform flow towards the lock gate. A weir at the other end is used to control the water level downstream of the lock gate.

The length of the flume allowed 20 m between the dissipating screen and the lock gate and 25 m downstream up to the weir.

2.3 Measurement set-up

During all tests the water level has been measured at eight locations, shown in Figure 1, using resistance water level gauges. Based on the variation in the signal an accuracy of 4 mm has been ascribed to these instruments.

Figure 1: Locations water level gauges in the flume. Top view; dimensions are given in mm.

The discharge through the pump has been

recorded. The accuracy of this value is 0.5%. Finally, a ray of five electromagnetic velocity

measurement instruments (EMS) is used at four different heights and four distances from the lock gate, as shown in Figure 2. This in total provides 80 locations of velocity measurements. These EMS instruments measure velocities in the horizontal plane.

The hydraulic condition was chosen such that both the water level difference and discharge could both be measured with high accuracy.

dissipating screen lock gate water level gauge





Figure 2: Locations of the EMS instruments, (a) top view and (b) side view. Dimensions are given in mm.

2.4 Scale model set-up

The scale model used in this study is not a representation of a specific lock gate design. The goal of the measurement allowed a somewhat simplified representation of the lock gate geometry. The tested lock gate is however based on typical lock gate dimensions. Used references include the Kleine Sluis and Zuidersluis of IJmuiden as well as the locks at Rozenburg, Hansweert and Schijndel; all in the Netherlands.

Figure 3 shows an illustration of the scale model design. The two openings in the door are rectangular and diverting downstream in both height and width. The dimensions of these openings are given in Table 1. The upstream width of the door openings combined is half of the gate width. The openings can be partially of completely opened by setting the valves. The valves are set manually and in a very accurate manner. The openings and valves have been made of PVC and aluminium respectively, to avoid any small deviation of the dimensions throughout the test series. For both the width and the set height of the openings, an accuracy of 0.5 mm has been assigned.

Figure 3: The scale model of the lock gate. Left: the upstream side, right: the downstream side.

At the downstream side of the doors breaking bars have been installed, four per gate opening. This type of breaking bars is often used in locks to distribute the flow through the door openings over a larger surface. The blockage by the bars in the model is 32%, which is common.

In order to obtain a realistic situation, it is desired to maintain a minimum water level downstream of the lock gate of 0.7 m model scale, as this equals a common lock chamber depth of 4 – 4.5 m. The upstream water level should be limited to 1.0 m, in which case the height of the openings are 20% of the water depth, which is a rule of thumb often used in lock gate design.

Table 1: Dimensions of the designed gates.

Parameter Model Proto-

type

Width of lock gate [m] 1.000 7.000

Height of lock gate [m] 1.200 8.400

Lock gate thickness [m] 0.150 1.000

Width of opening, inflow [m] 0.250 1.750

Height of opening, inflow [m] 0.200 1.250

Width of opening, outflow [m] 0.310 2.170

Height of opening, outflow [m] 0.250 1.750

Level of bottom edge of

opening from flume floor level [m] 0.100 0.700

Distance between breaking

bars [m] 0.042 0.300

Number of breaking bars [-] 4 4

Width of breaking bars [m] 0.025 0.175

Thickness of breaking bars [m] 0.025 0.175

Height of breaking bars [m] 0.300 0.210

Blocking fraction [%] 32 32

Thickness of the valve [m] 0.015 0.100

Width of the valve [m] 0.290 2.030

Height of the valve [m] 0.220 1.540

Thickness of the guiding rail [m] 0.015 0.100

Width of the guiding rail [m] 0.020 0.140

Height of the guiding rail [m] 0.500 3.500

2.5 Tested conditions

Three geometries were tested (T01, T02 and T03). Only one valve was opened in T01, with the other remaining shut completely. In test T02 both valves were used. Finally, in test T03, both openings were equipped with breaking bars and both were opened.

All three geometries were tested with four different opening heights: 10%, 25%, 50% and 100%.

The twelve tests are enumerated in Table 2. Figure 4 to Figure 7 show photographs taken of the valve openings during the various tests.

lock gate

EMS (a)

(b)





Table 2: Tested conditions.

Test

Number of

opened

valves

Breaking

bars

present

Opening

percentage

T01

a

1 No

10%

b 25%

c 50%

d 100%

T02

a

2 No

10%

b 25%

c 50%

d 100%

T03

a

2 Yes

10%

b 25%

c 50%

d 100%

Figure 4: T01a, looking from the upstream side. The left valve has been shut completely; the right valve is set at 10% of its opening height.

Figure 5: T01d, looking from the upstream side. The left valve has been shut completely; the right valve is set completely open.

Figure 6: T02a, looking from the downstream side. Both valves are set at 10% of their opening height.

Figure 7: T03d, looking from the upstream side. Both openings are equipped with breaking bars on the downstream side. Both valves are fully opened.

3 RESULTS SCALE MODEL

3.1 Discharge coefficients

The discharge coefficient can be computed from the measured water level difference and discharge. The water level gauges closest to the lock gate were used for this calculation. These were located at 4 m from either side of the lock gate.

Table 3 presents the measured discharge and water level difference for all twelve tests. Furthermore, the calculated discharge coefficient is given. The rightmost table presents the accuracy of the calculated discharge coefficient.

The change of the discharge coefficient as function of the relative valve opening is given in Figure 8. The highest discharge coefficients are found for the 10% opened valve. In this case, the dimensions of the opening are large in comparison to the valve opening, and the flow through the valve can diverge easily.





Table 3: Discharge coefficients resulting from measurements.

Test H B Δh Q μμ

[mm] [mm] [mm] [L/s] [-] [%]

T01a 20 250 666.7 14.4 0.798 5.21

T01b 50 250 328.8 22.2 0.700 2.90

T01c 100 250 292.0 40.4 0.676 2.43

T01d 200 250 189.3 68.6 0.713 3.20

T02a 20 500 374.6 21.2 0.783 5.24

T02b 50 500 274.8 40.7 0.702 3.06

T02c 100 500 199.8 68.2 0.689 3.17

T02d 200 500 67.4 84.8 0.740 8.60

T03a 20 500 405.4 21.8 0.773 5.29

T03b 50 500 292.2 41.4 0.692 2.96

T03c 100 500 211.8 68.6 0.673 3.04

T03d 200 500 84.9 85.9 0.667 6.81

Figure 8: Discharge coefficients as function of valve opening for the three tested geometries.

T01 and T02 show an increase of the discharge

coefficient when the valve is fully opened. An explanation for this can be found in the absence of flow separation at the valve’s tip. This flow separation does occur in case of partially opened valves, 25% and 50%, and induces a higher flow resistance.

T03 shows the lowest discharge coefficients or, equivalently, the highest flow resistance. This can be explained by the presence of breaking bars. In case of fully opened valve these bars are determining in the flow resistance, as can be expected.

The 2σ/µ value in Table 3 represents the 95%

confidence range. This has been established using a Monte Carlo analysis based on the accuracies of the geometry (surface A) and the measurements (water level difference Δh and discharge Q). The average accuracy is 4%, with maximum of 8.6%.

3.2 Velocities

Figure 9 and Figure 10 present the measured flow velocities for test T02 and T03 respectively. The four distances from the lock gate at which the velocities were measured are (from leftmost plot to rightmost plot) 0.50 m, 0.75 m, 1.00 m and 2.00 m.

Figure 9: Vertical flow velocity profiles for test T02 (a, b, c and d) Four locations along the flume (x-direction) were measured, and four locations in the vertical (z-direction). Five EMS instruments were positioned laterally (y-direction), shown in separate colours. The water level is indicated by the blue dotted horizontal line.

For T02, the maximum flow velocities can be

found at the level of the openings. A vertical recirculation pattern can be recognized although the flow back towards the lock gate has a relatively small magnitude in most cases.

The flow in the middle of the flume (y3, black line) is often very low. The two jets on either side

0.60

0.65

0.70

0.75

0.80

0.85

0% 20% 40% 60% 80% 100%

T01 T02 T03





seem roughly symmetric, with y2 (dark grey) and y4 (blue) giving comparable velocities.

Figure 10: Vertical flow velocity profiles for test T03 (a, b, c and d)

The measurement results for test T03 clearly

show the distribution of the flow over the vertical due to the breaking bars. Further away from the lock gate, e.g. compare x = 0.50 m and 0.75 m, the flow spreads more uniformly in the vertical.

The two filling jets are pronounced near the lock gate: velocities in the middle of the flume (black line) are close to zero in tests T03a, b and c. At x = 1.00 m and further, the flow is however distributed in the horizontal as well as the vertical.

4 SET-UP NUMERICAL MODEL

The CFD model is set up in the commercial software package Star-CCM+. It is a finite-volume code which solves the 3D Navier-Stokes equations of fluid motion. The turbulence of the flow is modelled using the Reynolds-Averaged Navier-Stokes (RANS) formulation. That entails that the mean velocity field is solved for and the effect of turbulent fluctuations on the mean field are accounted for by means of a turbulence model. Here the realizable k-ε model is used.

The geometry is generated with the Star-CCM+ 3D CAD module and the dimensions of the scale

model are used. The mesh is generated with the cell-splitting algorithm of Star’s Trimmer mesher. This creates hexahedral or rectangular cells. At least 10 cells are used across the narrowest opening of the gate valves and extra refinement of the mesh is added in the regions where the gradients of the velocity are highest, such as in the openings and in the downstream region of the jet. The total number of cells is approximately 1 million cells in all cases.

The free surface is model by means of a Volume of Fluid (VoF) method, whereby the air and water phases are signalled by a volume fraction variable, which is 1 in water and 0 in air. The location of the free surface is then indicated at the point where this variable is equal to 0.5. The interface should be no more than one computational cell thick. The boundary conditions are set such that all the walls are no-slip smooth walls and the upstream and downstream boundaries are set with constant water levels. The water levels are chosen to be consistent with the experiments.

An important difference with the experiments is that the water levels in the CFD are fixed and the discharge through the system is allowed to vary until equilibrium is reached. In the experiments the discharge and the downstream water level are fixed and the upstream water level allowed to vary until equilibrium. The results of the discharge coefficient are not affected by the small difference but the exact discharge and water levels may vary between the CFD and the experiment. The reason for using this approach in the CFD is simply for convenience of the numerical setup. If a mass flow is set at the upstream boundary it is set as a uniform flow across the inlet and a long upstream section is required to allow the flow to develop into an appropriate velocity profile. If just the water levels are set, the profile at the boundary can vary into an appropriate profile and therefore the boundary can be placed much closer to the door than in the other variant. This saves computational time.

5 RESULTS OF CFD

CFD computations focussing on the discharge coefficient have been done for all tests. The velocities from CFD computation presented in this paper are from simulations of tests T03.

Figure 11 shows the velocities in a cross-section for test T03c. This figure clearly shows the effect of the breaking bars on the flow field. Not only do the breaking bars distribute the flow over the opening’s height, the flow is effectively steered upward due to the presence of breaking bars. This has been seen in a comparable situation by O’Mahoney & De Loor (2015).





Figure 11: Vertical cross-section with breaking bars, door valve half open, test T03c.

The velocities of CFD will be discussed further in

Paragraph 6.2 in view of readability, combined with the comparison with measurements.

6 COMPARISON OF CFD AND MEASUREMENT RESULTS

6.1 Discharge coefficients

The water levels have been imposed as boundary conditions in the CFD computations. The water level difference, combined with the resulting discharge through the levelling openings provides the discharge coefficient according to these computations.

Table 3 shows the discharge coefficient from CFD computations (red) as well as from measurements (blue). The 95% confidence intervals on the measurement result values have been included in the graphs (black bars).

A very good agreement is shown, with a maximum difference between the discharge coefficient from CFD versus from measurements of 0.03, and the relative error not exceeding 5% in any case.

Figure 12a: Comparison of discharge coefficients from measurements (blue, with 95% confidence range) and CFD (red) for tests T01.

Figure 12b: Comparison of discharge coefficients from measurements and CFD for tests T02 and T03.

6.2 Velocity profiles

Figure 13 shows a comparison of the streamwise velocity; measurements shown as blue circles and CFD results shown as a red line. The sixteen rays of the EMS instruments are drawn: four locations along the flume and four locations in height.

In the proximity of the lock gate gate (x = 0.50 m and x = 0.75 m) the two filling jets could easily be distinguished from the measurements as has been discussed in Paragraph 3.2. The same pattern can be recognized in the velocity profiles following from CFD. Furthermore, the magnitudes of the velocities from CFD agree very well with those measured. The flow near the lock gate is heavily influenced by the geometry, including the breaking bars; it is important to note that this has been very accurately included in the 3D CAD geometry of the CFD.

Far away from the lock gate (x = 2.00 m), the flow has distributed over the flume width, showing from both CFD and measurements. The measured magnitudes of the velocities at this point agree well, although it can be seen that measured velocities are lower than those resulting from CFD computations.

At a distance of 1.00 m from the lock gate, the difference between the measurements and CFD is most striking. CFD shows the filling jets have fused to a great extent; the velocity profile is roughly bell shaped, with a recirculation along the flume walls. The measurements have not captured this bell





shape. Furthermore, since no EMS instruments were placed closer to the walls, the recirculation pattern does not show from measurements.

The streamwise development of the flow requires some distance: the character of the flow changes at a certain point from (fused) filling jets to a bell-shaped flow distribution. From the comparison of CFD and measurements at the ray 1.00 m from the lock gate it can be noted that this distance is not reproduced correctly in CFD. CFD predict a quicker dissipation of energy. The computations however capture the flow before and after this transition point very well.

Figure 13: Velocities (x-component) profiles in transversal direction, from measurement (blue) and CFD computation (red). Test T03c is shown: with breaking bars present and a half-opened valve.

7 CONCLUSIONS

The discharge coefficient of a levelling system is determinative for the discharge into or out of the lock during levelling. Consequentially, it determines the prediction of forces on ships during levelling, as these forces depend on the discharge.

As the discharge coefficient is very difficult to establish a priori and scale model tests are relatively expensive, especially early in the design process, CFD is being developed as an alternative. The measurements presented in this paper are intended to enable the validation of CFD for this purpose.

The measurements provide the discharge coefficient for three different geometries, based on typical lock gate designs. Four different valve openings are measured. Furthermore, velocities have been measured at eighty positions downstream of the gate.

CFD has shown to reproduce the discharge coefficients with great accuracy: the error made does not exceed 5% in any of the cases. Therefore, it can be concluded that CFD can be for the determination of discharge coefficients in case of lock gate openings.

The comparison of velocities from CFD and measurements shows a very good agreement in the proximity of the lock gate. Further downstream CFD predicts a quicker dissipation of energy than can be seen from measurements. Although an accurate representation of the flow field downstream of the lock gate is not necessary for a good prediction of the discharge coefficient, this is relevant if CFD would be used to compute forces on a ship in the lock chamber.

ACKNOWLEDGEMENTS

This project has been funded by the Dutch Ministry of Economic Affairs, within the project TO2 Natte Kunstwerken (Hydraulic Structures).

REFERENCES

LOCKFILL manual, version July 2015. Deltares, available through oss.deltares.nl

Loor, A. de, Weiler, O., Kortlever, W. (2013), LOCKFILL: A mathematical model for calculating forces on a ship while levelling through the lock head, Proceedings of the PIANC SMART Rivers Conference 2013, Paper 93

O’Mahoney, T.S.D., Loor, A. de (2015), Computational Fluid Dynamics simulations of the effects of density differences during the filling process in a sea lock, Proceedings of the PIANC SMART Rivers Conference 2015, Paper 55

Star-CCM+, Available from http://www.cd-adapco.com/products/star-ccm-plus


paper 105 comparison of scale model measurements and 3d

Documents